How to Trace Your First Agent

You will instrument an OpenAI LLM call with HoneyHive and view the full request and response in the dashboard in about 5 minutes. This quickstart uses automatic tracing via OpenInference; for production web servers or Lambda, see tracer initialization after you finish. Expected time: 5 minutes

Create a project

Go to HoneyHive Projects and click New Project. Name it my-first-project.

Get your API key

Go to Settings > Project > API Keys. The page opens on the Ingestion tab. Click Create API Key, name the key, and click Create Key. Copy the key from the dialog and store it as HH_INGESTION_API_KEY - it is only shown once.An ingestion key can only send traces and events. To run experiments or read data back, also create a project API key on the Project tab and store it as HH_API_KEY. The SDK uses each key where it applies. See Ingestion keys.

Install dependencies

pip install honeyhive openai openinference-instrumentation-openai

Requires Python 3.11+.

Using a different provider? See our integration guides for Anthropic, Google, AWS Bedrock, and more.

Run this script

Create a new file called quickstart.py and paste the following code. It reads your ingestion API key from HH_INGESTION_API_KEY, so export that first (API keys are project-scoped).

from honeyhive import HoneyHiveTracer
from openinference.instrumentation.openai import OpenAIInstrumentor
from openai import OpenAI

# Initialize HoneyHive tracing
tracer = HoneyHiveTracer.init(
    source="development",  # Optional: Label where traces come from
    session_name="quickstart-session",  # Optional: Group related traces
    # server_url="https://your-instance.honeyhive.ai"  # Required for self-hosted/dedicated
)

# Add the OpenAI instrumentor
OpenAIInstrumentor().instrument(tracer_provider=tracer.provider)

# Make a traced OpenAI call
client = OpenAI()
response = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": "Write a fun haiku about honeybees"}]
)

print(response.choices[0].message.content)

This quickstart keeps tracing setup simple: initialize the tracer first, then initialize the instrumentor with tracer.provider.

To see where to initialize the tracer for your environment, including AWS Lambda and long-running servers, see Tracer Initialization.

Run the script:

python quickstart.py

View your trace

Go to the Traces page in HoneyHive. Click on your trace to see the full request and response.

What should you do next?

Now that you’re tracing, try the Experiments Quickstart to run your first evaluation and compare prompt variations. To understand how HoneyHive structures sessions and events, read tracing concepts. To trace your own business logic beyond LLM calls, see custom spans.

Choose runtime setup

See where to initialize the tracer for Lambda, web servers, and script-style apps.

Add more providers

Trace Anthropic, Google, AWS Bedrock, and more.

Trace custom functions

Use decorators to trace your own business logic.

Enrich your traces

Add feedback, metadata, and user information.

Run evaluations

Set up automated evaluations on your traces.

Getting Started

Observability

Evaluation

Prompt Management

Administration

Learn More

What should you do next?

Choose runtime setup

Add more providers

Trace custom functions

Enrich your traces

Run evaluations

​What should you do next?

Choose runtime setup

Add more providers

Trace custom functions

Enrich your traces

Run evaluations

What should you do next?